Automatic backlit face detection

Abstract

The subject application is directed to a system and method for detecting facial areas from an image having a relatively light background. Image data is received including at least one candidate facial region. A candidate facial region is isolated from the image data as is a subportion of an interior of the candidate facial region. A size of the subportion is compared relative to a preselected threshold size value, a luminance value of the subportion is compared with a preselected darkness threshold value, and a facial recognition signal is generated corresponding to a detected facial region in accordance with an output of the size comparison and the luminance comparison.

1 . A system for detecting facial areas from images having a relatively light background comprising:
 an input operable to receive image data including at least one candidate facial region;   an isolator operable to isolate a candidate facial region from received image data;   an image segmentor operable to isolate a subportion of an interior of the candidate facial region;   a size comparator operable to compare a size of the subportion relative to a preselected threshold size value;   a darkness comparator operable to compare a luminance value of the subportion with a preselected darkness threshold value; and   a facial region signal output operable to output a signal corresponding to a detected facial region in accordance with an output of the size comparator and the darkness comparator.   
     
     
         2 . The system of  claim 1  further comprising:
 a histogram generator operable to generate luminance histogram data of the subportion;   a normalizer operable to normalize the luminance histogram data; and   wherein the darkness comparator is further operable to compare normalized luminance histogram data with the preselected darkness threshold value.   
     
     
         3 . The system of  claim 2  further comprising:
 a normalized histogram accumulator operable on the normalized histogram data;   a mid-point isolation system operable to isolate a mid-point of the normalized histogram data; and   wherein the darkness comparator is further operable to compare the mid-point to the preselected darkness threshold value.   
     
     
         4 . The system of  claim 3  further comprising:
 a threshold size value generator operable to generate the threshold size value in accordance with a plurality of previous facial candidate portion measurements; and   a darkness threshold value generator operable to generate the darkness threshold value in accordance with the plurality of previous facial candidate portion measurements.   
     
     
         5 . The system of  claim 4  further comprising a correction signal generator operable to generate an image correction signal in accordance with an output of at least one of the size comparator and the darkness comparator. 
     
     
         6 . The system of  claim 5  further comprising an image adjustor operable to generate adjusted image data corresponding to received image data in accordance with the image correction signal. 
     
     
         7 . A method for detecting facial areas from images having a relatively light background comprising the steps of:
 receiving image data including at least one candidate facial region;   isolating a candidate facial region from received image data;   isolating a subportion of an interior of the candidate facial region;   comparing a size of the subportion relative to a preselected threshold size value;   comparing a luminance value of the subportion with a preselected darkness threshold value; and   generating a facial recognition signal corresponding to a detected facial region in accordance with an output of the size comparison and the luminance comparison.   
     
     
         8 . The method of  claim 7  further comprising the steps of:
 generating histogram data of the subportion;   normalizing the luminance histogram data;   comparing normalized luminance histogram data with the preselected darkness threshold value.   
     
     
         9 . The method of  claim 8  further comprising the steps of:
 generating normalized histogram data from the histogram data;   isolating a mid-point of the normalized histogram data; and   comparing the mid-point to the preselected darkness threshold value.   
     
     
         10 . The method of  claim 9  further comprising the steps of:
 generating the threshold size value in accordance with a plurality of previous facial candidate portion measurements; and   generating the darkness threshold value in accordance with the plurality of previous facial candidate portion measurements.   
     
     
         11 . The method of  claim 10  further comprising generating an image correction signal in accordance with an output of at least one of the size comparison and the darkness comparison. 
     
     
         12 . The method  claim 11  further comprising generating adjusted image data corresponding to received image data in accordance with the image correction signal. 
     
     
         13 . A system for detecting facial areas from images having a relatively light background comprising:
 means adapted for receiving image data including at least one candidate facial region;   means adapted for isolating a candidate facial region from received image data;   means adapted for isolating a subportion of an interior of the candidate facial region;   size comparison means adapted for comparing a size of the subportion relative to a preselected threshold size value;   a luminance comparison means adapted for comparing a luminance value of the subportion with a preselected darkness threshold value; and   means adapted for generating a facial recognition signal corresponding to a detected facial region in accordance with an output of the size comparison means and the luminance comparison means.   
     
     
         14 . The system of  claim 13  further comprising:
 a histogram generator means adapted for generating histogram data of the subportion;   means adapted for normalizing the luminance histogram data; and   wherein the luminance comparison means includes means adapted for comparing normalized luminance histogram data with the preselected darkness threshold value.   
     
     
         15 . The system of  claim 14  further comprising:
 means adapted for generating normalized histogram data from the histogram data;   means adapted for isolating a mid-point of the normalized histogram data; and   wherein the luminance comparison means includes means adapted for comparing the mid-point to the preselected darkness threshold value.   
     
     
         16 . The system of  claim 15  further comprising:
 a threshold size value generator means adapted for generating the threshold size value in accordance with a plurality of previous facial candidate portion measurements; and   a darkness threshold generator means adapted for generating the darkness threshold value in accordance with the plurality of previous facial candidate portion measurements.   
     
     
         17 . The system of  claim 16  further comprising a correction signal generator means adapted for generating an image correction signal in accordance with an output of at least one of the size comparison means and the darkness comparison means. 
     
     
         18 . The system of  claim 17  further comprising an image adjustor means adapted for generating adjusted image data corresponding to received image data in accordance with the image correction signal.